Clustering Ammar's Splicing Dataset¶

Author: Shehbeel Arif¶

Preclinical Laboratory Research Unit¶

The Center for Data Driven Discovery in Biomedicine (D3b)¶

Children's Hospital of Philadelphia¶

In [1]:
# Library for data handling
import pandas as pd

# K-means classification
from sklearn.cluster import KMeans

# Library for UMAP projections
from umap.umap_ import UMAP
# Library for t-SNE projects
from sklearn.manifold import TSNE
# PCA clustering
from sklearn.decomposition import PCA

# Library for visualization
import matplotlib.pyplot as plt
import plotly.express as px
In [2]:
df = pd.read_csv('pan_cancer_splicing_SE_gene.txt', sep='\t')
df
Out[2]:
Splice_ID BS_15ETQ0E4 BS_1QNTEZPS BS_31TF0P1N BS_4E1G7E3G BS_6G6C7VPX BS_850BAHH9 BS_90B22NGP BS_996ZS2A4 BS_9FHSNVS0 ... BS_WH5HDE0A BS_Y8D1F5WW BS_YEB88D32 BS_YETBN0RQ BS_Z45HD492 BS_Z7PKVY9J BS_ZV68CES9 BS_ZXYDSBM9 BS_BKR4E30C BS_C7N3J4NW
0 CLIC2 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 ... 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
1 FAM3A 0.000 0.821 0.300 0.852 0.759 0.776 0.794 0.704 0.320 ... 0.889 0.591 0.000 0.786 0.868 0.000 0.816 0.753 0.814 0.846
2 MECP2 0.502 0.612 0.434 0.512 0.601 0.615 0.498 0.622 0.445 ... 0.558 0.509 0.629 0.430 0.558 0.750 0.462 0.597 0.518 0.480
3 PDZD4 0.000 0.000 0.000 0.000 0.000 0.000 0.882 0.000 0.894 ... 0.791 0.000 0.000 0.000 0.762 0.796 0.894 0.133 0.135 0.630
4 MTMR1 0.000 0.245 0.000 0.000 0.169 0.000 0.000 0.000 0.000 ... 0.146 0.541 0.000 0.253 0.764 0.000 0.362 0.319 0.000 0.000
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
14299 XKRX 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 ... 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
14300 GSTM5 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 ... 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
14301 TLR8 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 ... 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
14302 SEL1L2 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 ... 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
14303 NYAP1 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 ... 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

14304 rows × 604 columns

UMAP Plot¶

In [4]:
umap_2d = UMAP(n_components=2, init='random', random_state=0)
#umap_3d = UMAP(n_components=3, init='random', random_state=0)

proj_2d = umap_2d.fit_transform(df.iloc[:,1:])
#proj_3d = umap_3d.fit_transform(new_df.loc[:, new_df.columns != 'class'])

fig_2d = px.scatter(proj_2d, x=0, y=1)

#fig_3d = px.scatter_3d(proj_3d, x=0, y=1, z=2, color=new_df['class'].tolist())

fig_2d.show()
#fig_3d.show()

PCA Plot¶

In [5]:
# Initialize PCA object
pca = PCA(n_components=2) # Looking at only the first four principal components

# Perform PCA on PBTA dataset
pca_comps = pca.fit_transform(df.iloc[:,1:])

# Perform PCA on PBTA dataset
pca_comps = pca.fit_transform(df.iloc[:,1:])

# Display PCA plot based on K-means Classification
fig = px.scatter(pca_comps, x=0, y=1)

fig.show()

t-SNE Plot¶

In [6]:
# Initialize t-SNE object
tsne_2d = TSNE(n_components=2, random_state=0, learning_rate='auto')
tsne_3d = TSNE(n_components=3, random_state=0, learning_rate='auto')

# Fit and transform the data
tsne_proj_2d = tsne_2d.fit_transform(df.iloc[:,1:])

fig_2d = px.scatter(tsne_proj_2d, x=0, y=1)

fig_2d.show()

/Users/arifs2/opt/anaconda3/lib/python3.9/site-packages/sklearn/manifold/_t_sne.py:780: FutureWarning:

The default initialization in TSNE will change from 'random' to 'pca' in 1.2.

Unsupervised Clustering using K-means¶

In [7]:
# Finding optimal k in K-means
Sum_of_squared_distances = []
K = range(1,10)
for num_clusters in K :
 _kmeans = KMeans(n_clusters=num_clusters)
 _kmeans.fit(df.iloc[:,1:])
 Sum_of_squared_distances.append(_kmeans.inertia_)
plt.plot(K,Sum_of_squared_distances,"bx-")
plt.xlabel("Values of K") 
plt.ylabel("Sum of squared distances/Inertia") 
plt.title("Elbow Method For Optimal k")
plt.show()
In [10]:
# K-means classification k=2
kmeans = KMeans(n_clusters=2, random_state=0).fit(df.iloc[:,1:])
kmeans.labels_

umap_2d = UMAP(n_components=2, init='random', random_state=0)

proj_2d = umap_2d.fit_transform(df.iloc[:,1:])

fig_2d = px.scatter(proj_2d, x=0, y=1, color=kmeans.labels_)

fig_2d.show()

Clustering by Splicing Sites¶

In [11]:
dft = df.T
dft.columns = dft.iloc[0]
dft = dft.drop(dft.index[0])
dft
Out[11]:
Splice_ID CLIC2 FAM3A MECP2 PDZD4 MTMR1 SLITRK2 LINC00632 STAG2 THOC2 GRIA3 ... FAT2 PMEL PERM1 IRF8 TPSG1 XKRX GSTM5 TLR8 SEL1L2 NYAP1
BS_15ETQ0E4 0.0 0.0 0.502 0.0 0.0 0.0 0.0 0.245 0.144 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
BS_1QNTEZPS 0.0 0.821 0.612 0.0 0.245 0.0 0.0 0.54 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
BS_31TF0P1N 0.0 0.3 0.434 0.0 0.0 0.0 0.0 0.43 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
BS_4E1G7E3G 0.0 0.852 0.512 0.0 0.0 0.0 0.0 0.213 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
BS_6G6C7VPX 0.0 0.759 0.601 0.0 0.169 0.0 0.0 0.738 0.16 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
BS_Z7PKVY9J 0.0 0.0 0.75 0.796 0.0 0.0 0.0 0.543 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
BS_ZV68CES9 0.0 0.816 0.462 0.894 0.362 0.0 0.0 0.756 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
BS_ZXYDSBM9 0.0 0.753 0.597 0.133 0.319 0.0 0.0 0.644 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
BS_BKR4E30C 0.0 0.814 0.518 0.135 0.0 0.598 0.0 0.852 0.0 0.141 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
BS_C7N3J4NW 0.0 0.846 0.48 0.63 0.0 0.841 0.5 0.457 0.0 0.752 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

603 rows × 14304 columns

In [12]:
umap_2d = UMAP(n_components=2, init='random', random_state=0)

proj_2d = umap_2d.fit_transform(dft)

fig_2d = px.scatter(proj_2d, x=0, y=1)

fig_2d.show()
In [14]:
# Finding optimal k in K-means
Sum_of_squared_distances = []
K = range(1,10)
for num_clusters in K :
 _kmeans = KMeans(n_clusters=num_clusters)
 _kmeans.fit(dft)
 Sum_of_squared_distances.append(_kmeans.inertia_)
plt.plot(K,Sum_of_squared_distances,"bx-")
plt.xlabel("Values of K") 
plt.ylabel("Sum of squared distances/Inertia") 
plt.title("Elbow Method For Optimal k")
plt.show()
In [15]:
# K-means classification k=3 and 4
kmeans = KMeans(n_clusters=3, random_state=0).fit(dft)
#kmeans.labels_

umap_2d = UMAP(n_components=2, init='random', random_state=0)

proj_2d = umap_2d.fit_transform(dft)

fig_2d = px.scatter(proj_2d, x=0, y=1, color=kmeans.labels_)

fig_2d.show()